Gemini 4 Hydrogen vs Carbon vs Argon: Price, Capability and Which to Pick

Short answer

Gemini 4 now has three models: Hydrogen, Carbon and Argon. They share one architecture — the same 1M unified context window, the same long-decode behaviour, the same built-in tools. The only real variable is how many thinking tokens each burns per turn, meaning how long it deliberates.

Which to pick:

• Everyday coding, agents, multimodal documents → Hydrogen (gemini-4-hydrogen), the cheapest tier at $1.50 / $7.50.
• Hard architecture, subtle bugs, science, security audits → Carbon (gemini-4-carbon), strongest but priciest at intro $2 / $10.
• Long-horizon work without burning tokens → Argon (gemini-4-argon), in between, also intro $2 / $10.

TeamoRouter carries Argon and Carbon today; Hydrogen is not listed yet. Once it is, the same key and a new model ID is all it takes.

The whole comparison in one table

Where they actually differ

Google's own framing splits the work three ways:

Hydrogen — cheap and fast, but not dumb. Google describes it as bringing Gemini 4 intelligence down to a cheaper, faster tier, and explicitly names it the recommended upgrade from Gemini 3.8 Flash and earlier Gemini 3 and 2.5 models. The pitch: cleaner code, and when context is missing it says so instead of inventing it (Google claims more than 5x fewer fabricated claims than 3.8 Flash), with tool calls batched in parallel. Built for everyday coding, interactive web apps, tool-using agents, multimodal documents and parallel subagents.

Carbon — strongest, deepest reasoning. Google calls it its strongest frontier model. It spends more thinking tokens per turn testing hypotheses and verifying edge cases. Built for hard architecture, debugging, scientific reasoning and security audits. Google's published cybersecurity benchmarks (finding, validating and patching vulnerabilities) are Carbon's territory.

Argon — handles most long tasks with fewer tokens. Google positions it as handling the majority of advanced long-horizon tasks with lower token use than Carbon. It shares Carbon's intro price, so the reason to pick it is not savings — it is that the task does not need Carbon's depth.

The counter-intuitive part of the pricing

Carbon and Argon cost the same during the intro period ($2 / $10), and Hydrogen is cheaper ($1.50 / $7.50).

That means:

• If both Carbon and Argon can do the job, pick Argon — same money, less deliberation, faster.
• If the job does not need Carbon-grade reasoning, Hydrogen saves you a third.
• But cheaper is not automatically cheaper overall. Thinking and output tokens bill together, so if Hydrogen's shallower reasoning makes you retry twice, the total can exceed a single Carbon call.

A practical way to choose

Work down this list; do not start with the most expensive model:

1. Run a batch of your real requests through Hydrogen at the default high thinking level. Do not change anything yet.
2. Check whether the output is good enough. If it is, stop — this is the cheapest option.
3. If not, try Hydrogen at xhigh. Same model, more deliberation on hard problems. Be aware xhigh over-thinks simple prompts, so use it selectively.
4. If xhigh is still not enough, move to Carbon. At that point the task genuinely needs the strongest reasoning.
5. If it is long-horizon work and Carbon is too expensive, try Argon. It spends fewer tokens than Carbon and covers most long tasks.

Google also offers a pairing worth copying: have Carbon plan system interfaces and clear hard roadblocks, then hand code edits, parallel subagent tasks and test runs to Hydrogen. Put the expensive model where decisions are made and the cheap one where work is executed. In an agent loop that saves real money.

Against Gemini 3.8 Flash

Hydrogen takes over 3.8 Flash's slot. On price:

Right now — through December 31, 2026 — 3.8 Flash is exactly half of Hydrogen's price. So in the short term:

• High-volume, simple calls → 3.8 Flash is still the better deal.
• Work that needs stronger coding and agentic behaviour → Hydrogen, and the extra spend buys fewer retries.
• On January 1, 2027, 3.8 Flash rises to Hydrogen's price, and there is no reason to keep using it. Hydrogen takes over completely.

That date matters. Put the migration on the calendar now.

Calling them on TeamoRouter

The two available today:

from openai import OpenAI

client = OpenAI(base_url="https://api.teamorouter.com/v1", api_key="YOUR_KEY")
resp = client.chat.completions.create(
    model="gemini-4-carbon",  # or gemini-4-argon
    messages=[{"role": "user", "content": "..."}],
)
print(resp.choices[0].message.content)

gemini-4-hydrogen is not on TeamoRouter yet. Once it is live, the same key and model: gemini-4-hydrogen is all you change; until then, use gemini-4-argon.

How to pay

No subscription: top up your balance in the console from $5, billed by the tokens each request uses. Cards are supported, and so are USDT, USDC and ETH. The same key also calls Claude and GPT models.

New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.

FAQ

What actually separates the three models?

The architecture and context window are identical. What differs is how many thinking tokens each spends per turn: Hydrogen least (fast, cheap), Argon in the middle, Carbon most (strongest, priciest).

Which is cheapest?

Hydrogen: $1.50 / $7.50. Carbon and Argon are both $2 / $10 during the intro period, rising to $4 / $20 after November 30, 2026.

Argon costs the same as Carbon — why pick Argon?

Because it spends fewer tokens. Same price, but faster and lighter deliberation, which suits tasks that are not that hard. Carbon costs more precisely because it re-verifies across more thinking tokens per turn.

Can Hydrogen replace Carbon?

Not entirely. Carbon is stronger on hard architecture, subtle bugs and security audits. But Hydrogen at high or xhigh covers most everyday work.

Which should I use right now?

TeamoRouter carries Argon and Carbon today: Argon for everyday work, Carbon for hard problems. Once Hydrogen is listed, try it first — it is the cheapest.

What happens on January 1, 2027?

Gemini 3.8 Flash rises to $1.50 / $7.50, matching Hydrogen. At that point there is no reason to stay on 3.8 Flash; migrate to Hydrogen.

Related reading

• Gemini 4 Hydrogen API is out: price, the four thinking levels and how to call it
• Using Gemini 4 Hydrogen without a Google account: access from China and Russia
• Gemini 4 Carbon API: the strongest tier
• Gemini 4 Argon API: the middle tier available today